Statistical Patterns in Natural Lightning Data

Every day millions of lightning flashes occur around the globe but the understanding of this natural phenomenon is still lacking, mostly due to the wide range of timescales involved and to the difficulty of obtaining measurements inside storms. Fundamentally, lightning is nature's way of destroying charge separation in clouds and restoring electric neutrality. Thus, statistical patterns of lightning activity indicate the scope of these electric discharges and offer a surrogate measure of timescales for charge buildup in thunderclouds. We develop a statistical method to investigate spatial and temporal correlations among lightning flashes using National Lightning Detection Network (NLDN) geo-location data. We present two sets of results, the first involving the patterns of flashes in a storm, and the second involving the patterns of subsequent return strokes in cloud-to-ground (CG) flashes. First, by monitoring the distribution of lightning activity, we can observe the charging and discharging processes in a given thunderstorm. We find that, following a lightning flash, the probability of another neighboring flash decreases and recovers with time. The probability of another flash within 10 km is reduced and takes tens of seconds to recover to the pre-flash value. This suppression effect is potentially a function of variables such as storm location, storm phase, and other lightning parameters. Second, we use NLDN data to study the evolution of the return strokes within a CG flash. A CG flash typically includes multiple return strokes, which can occur in the same channel or in multiple channels within a few kilometers. We cluster NLDN stroke data into flashes and produce the probability density function of subsequent strokes as a function of distance and time-delays relative to the previous stroke. Using this method, we investigate processes which occur during the CG lightning flash with nanosecond to millisecond timescales. For example, our results suggest that subsequent strokes that occur in a newly formed channel follow a pattern that propagates at a speed of ~200 km/s. We present our statistical techniques and discuss more thoroughly our work and results.